VOGONS


Reply 40 of 73, by mario990

User metadata
Rank Newbie
Rank
Newbie

I will replace FBI 500-0009-01 systems with Voodoo Diamond 3D II and ICUVGA-GW803 cards for the test.
Diamond 3D II has already soldered the new FBI 500-0009-01 I wrote about it in my first post, but I will exchange it and see what will happen ...

There is one more thing that puzzles me, TMU0 and TMU1 500-0010-01 systems were still dead when they were mounted on Diamond 3D II,
which may indicate a lack of power supply or connection with the FBI, you can see it here.
I did the test if the FBI Diamond 3DII correctly recognizes RAM with a full 4MB bank and 2MB desoldering, as you can see it recognizes correctly.

DIAMOND 3D II log not ok 4MB FBI.jpg
Filename
DIAMOND 3D II log not ok 4MB FBI.jpg
File size
269.73 KiB
Views
2034 views
File license
Fair use/fair dealing exception
DIAMOND 3D II log not ok 2MB FBI.jpg
Filename
DIAMOND 3D II log not ok 2MB FBI.jpg
File size
168.15 KiB
Views
2034 views
File license
Fair use/fair dealing exception

The log below ICUVGA-GW803 is good, despite the fact that TMU0 and TMU1 500-0010-01 had one leg broken off.
This leg connected with the memory, so the log is correct and the game Need For Speed III crashed due to a problem with texturing.

ICUVGA-GW003 log ok.jpg
Filename
ICUVGA-GW003 log ok.jpg
File size
233.64 KiB
Views
2034 views
File license
Fair use/fair dealing exception

Reply 43 of 73, by mario990

User metadata
Rank Newbie
Rank
Newbie

The conversion of 500-0009-01 processors between Voodoo has changed nothing.
Diamond 3D II shows the same fault when starting "_GlideInitEnvironment ..."
ICUVGA-GW803 is working properly, you can start the game Need For Speed III, there is a normal image, now it displays vertical stripes, 500-0009-01 has overheated. Once I had a similar situation with Voodoo 1, the 500-0004-02 TMU system overheated.
3Dfx processors do not like high temperature during soldering.

Your card probably also has 500-0009-01 good, the fault must be on the side of the board. I measured the AS2830 stabilizer voltages, are you Vin = 5V; Vout = 3.48V; ADJ = 2.27V, looks approx.

Reply 44 of 73, by feipoa

User metadata
Rank l33t++
Rank
l33t++

It could be soldering temperature, but as I touched each pin individually with a fine iron tip for only a second, I'd be surprised. I touched the pins just long enough for the solder to melt. I messed up pins on the second reflow because I thought I'd try to push the leads of the QFP in just a smidge to make the leads closer to the pads. Well, some leads slipped in.

Possibly where the solder pad meats a middle layer trace there is a disconnect?

Plan your life wisely, you'll be dead before you know it.

Reply 45 of 73, by mario990

User metadata
Rank Newbie
Rank
Newbie

Today I measured the paths between the FBI 500-0009-01 and the PCI slot as well as the FBI and GENDAC ICS5342 - I found no damage.
I replaced GENDAC and (QS5244 diamond) (74AC244 ICUVGA-GW803) at the same time.
Voodoo ICUVGA-GW803 works ok, starts the game.
Diamond 3D II - FBI could not recognize GENDAC by 74AC244, the problem was the resistor R148 - there is a difference in value between the cards.
After changing the resistor, Diamond sees GENDAC, but the error "_GlideInitEnvironment ..."
When measuring the resistance on a 5V power line, I noticed the difference between the Voodoo Diamond and GW803 test.
The measurement was confirmed by two other Voodoo 2 CT6670 2455P which were complete with all components and 100% functional.
By measuring the ICUVGA-803 resistance around 200k between ground and 5V on C45 C4 C17 C35 C37 C9 capacitors and the AS2830 stabilizer, you can see how the resistance gradually increases beyond the set range of 10k, 23k, 32k, 45k, 70k, 90k, 120k, 168k 185k ... which indicates charging capacitors.
CT6670 2455P behaves in the same way.

Diamond II 3D behaves completely differently, it is small you can measure on the 2k range and the resistance decreases 137ohm, 68ohm, 40ohm, 18ohm. The voltage I measured is good Vin = 5v Vout 3.48V

feipoa, you can check how your Diamond 3D II behaves, what do you think about it?

I insert a photo Diamond 3D II and ICUVGA-GW803 without 500-0009-01 and 500-0010-01 processors, maybe someone will need to see the path layout under the processor.

Diamond 3D II.jpg
Filename
Diamond 3D II.jpg
File size
1.01 MiB
Views
1981 views
File license
Fair use/fair dealing exception
ICUVGA-GW803.jpg
Filename
ICUVGA-GW803.jpg
File size
1.73 MiB
Views
1981 views
File license
Fair use/fair dealing exception

Reply 46 of 73, by mario990

User metadata
Rank Newbie
Rank
Newbie

Diamond 3D II looks like the capacitors on the 5V line want to charge and one of them had a short circuit
(or something else on the board) and immediately discharged them. It is strange that after starting the computer is V = 5V without any drop.
On each capacitor on the 5V line I measure only 12ohm, the resistance is constant does not increase.
I disconnected the 5V AS2830 leg, it did not help, the resistance is still there, the stabilizer is in working order.

Reply 47 of 73, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Both of my computer test beds are in use right now. I have been testing HWiNFO and CPU-Z. With luck, one will free up this evening. What did you want me to measure, the power-off resistance across all the caps which go from 5V to GND?

Plan your life wisely, you'll be dead before you know it.

Reply 48 of 73, by mario990

User metadata
Rank Newbie
Rank
Newbie

Measure the resistance on the 5V line as I drew on the example, you can measure in this place or on any capacitor C42, C40, C43, C45, C37, C17, C3, C4 .... as you find it more convenient.
Set the meter to 200k ohm range and measure.
The correct measurement is to look that the resistance will gradually increase 10k, 23k, 32k, 45k, 70k, 90k, 120k, 168k 185k as far out of range,
this indicates the correct charging of the capacitors and that the capacitors and their circuit are not short-circuited.
I had a measurement of 12 ohm nothing increased.
To find the damage I had to remove most of the capacitors on the 5V line because measuring the capacity with a meter on the board was impossible and check everyone on the table.
I found a fault - the C28 capacitor was broken, its measurement showed a resistance of R = 12 ohm, the capacitor had no capacity, it did not charge - it was damaged.
Now after replacing with a new measurement it is correct and the measured resistance increases.
However, this issue did not resolve the problem, the error "_GlideInitEnvironment: glide 2x.dll expected Voodoo graphics, none detectedt ..." still appears.
I am still looking for damage.

Attachments

  • Diamond.jpg
    Filename
    Diamond.jpg
    File size
    1002.33 KiB
    Views
    1950 views
    File license
    Fair use/fair dealing exception

Reply 50 of 73, by gex85

User metadata
Rank Member
Rank
Member

Sorry for hijacking, but I have a Diamond Voodoo2 card with a missing C128 capacitor (backside, upper left corner, next to the memory chip). Does anyone happen to know which value it should be? I could of course de-solder one of the other SMD caps that are sitting next to the memory chips, but I'd rather not do it if it's not absolutely necessary. Thanks!

My retro computers

Reply 51 of 73, by trozx

User metadata
Rank Newbie
Rank
Newbie

Hello!

I'm grateful that this thread exists.

Unfortunately I'm in the same fail boat here.

I stored 2 working Diamond M3D II v 1.0s for about 15 years, in a static bag, always careful, and thinking I would use them again in the future.

So, I got the idea to build a classic rig for some fun a few days ago. After putting everything together, I get the dreaded "Not detected". Even though it is correctly configured in device manager. Both cards are acting exactly the same way, together in SLI or individually. Different slots, etc.

It's disheartening really. Perfectly working boards, well cared for, and they have low hours. Odd that they both failed the same way. I begin to think some passive components have gone out of spec after 20 years.

Going to give them a very good cleaning, remove stickers, etc. Then a bit of deoxit on the gold fingers. I might eventually reflow some pins, we'll see.

Other thoughts: All these cases of nearly the same failure. It also got me thinking, the one thing I don't have is the same vintage power supply for the rig. I wonder if anyone does? Perhaps these boards need exactly the original ATX spec and not backwards compatibility?

Finally, has anyone tracked down the original engineers at 3DFX? Maybe they would lend some insight!

I'll update again after trying some things.

Reply 52 of 73, by Thermalwrong

User metadata
Rank Oldbie
Rank
Oldbie

If they were put away working, they've been stored well and don't have damaged pins, they should be fine. Clean the PCI slot pins using cotton buds and IPA, oxidation does cause problems like that - I had a Gravis Ultrasound clone that was very patchy on detection until I did that. Do that before reflowing or trying to replace components at least.

Also try a couple of different motherboards, maybe a Slot 1 motherboard to see if it helps.

Flexing of the card could cause these tough to track down faults. Maybe if the two parts were squished in a box.
The other day I fixed an LCD panel that would just display noise unless you pressed it on the back in the right spot. Rather than track down the fault, I was able to repair it by resoldering all capacitors, resistors and IC's pads under a microscope with flux, then cleaned up. I did this because I didn't want to apply lots of heat to the area, it's a delicate PCB bordering a heat sensitive LCD panel.
These 90s PCBs are a similar problem, too much heat and you'll have pads falling off, as I can attest to. That makes the repair far more difficult.

Given that there are 3 main ICs, each with lots of pins, you can try this method demonstrated by 'scorp' of this forum, to track down loose pins: https://www.youtube.com/watch?v=VUx7jZWkYQc

Reply 53 of 73, by TSG

User metadata
Rank Newbie
Rank
Newbie

One of my STB Voodoo2 cards has started giving me trouble. I had them in SLI with no problems when I had the P3 450. After swapping in the P3 800, I was playing Q3 for about 15mins when the system froze on a grey screen with "Excellent" playing on repeat.

Before rebooting, I checked the temp of those cards and they were getting pretty warm. I zip-tied a fan in one of the case vents to get some more airflow going.

When I turned it back on, the second card wasn't being found. SLI was listing "Not Detected". So I tried both cards by themselves and found that one of them consistently crashes on Q3 after 15min. Grey screen, sounds looping, and locked up.

I hope I haven't thermally damaged it. I know absolutely nothing about soldering or troubleshooting individual chips.

Reply 54 of 73, by trozx

User metadata
Rank Newbie
Rank
Newbie

Quick update. I just reflowed all IC pins on one of my voodoo2s with success. The card is properly detected and plays games, albeit the colors are not right. Might have to replace output selector and ramdac, after testing, per this thread:

Successfully fixed a 3DFx Voodoo2 with miscolored output

Going to try reflowing the 2nd one as well.

Mojo.jpg
Filename
Mojo.jpg
File size
1.53 MiB
Views
1594 views
File license
Public domain
Tomb1.jpg
Filename
Tomb1.jpg
File size
1.18 MiB
Views
1594 views
File license
Public domain
Tomb2.jpg
Filename
Tomb2.jpg
File size
1.41 MiB
Views
1594 views
File license
Public domain

Edit:

Initially I was testing the V2 in my DOS Socket 7 box (Pentium 133) - Weird colors in games.

On a whim, I put it in my P3 Win95 box. Works perfectly.

OK COOL, but what's up with my Socket 7 machine? Maybe I need to set my variables.

Basically it looks like reflowing the card got it working 100% - I'll take it!

Last edited by trozx on 2021-11-22, 02:32. Edited 1 time in total.

Reply 55 of 73, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie
trozx wrote on 2021-11-21, 23:19:
Quick update. I just reflowed all IC pins on one of my voodoo2s with success. The card is properly detected and plays games, alb […]
Show full quote

Quick update. I just reflowed all IC pins on one of my voodoo2s with success. The card is properly detected and plays games, albeit the colors are not right. Might have to replace output selector and ramdac, after testing, per this thread:

Successfully fixed a 3DFx Voodoo2 with miscolored output

Going to try reflowing the 2nd one as well.

Mojo.jpg
Tomb1.jpg
Tomb2.jpg

That's great news, thanks for the update.

I might give this a go on my faulty STB Voodoo 2 that is no longer detected.

Care to elaborate on the process you used? Did you add any extra solder?

If you are squeamish, don't prod the beach rubble.

Reply 56 of 73, by trozx

User metadata
Rank Newbie
Rank
Newbie
DonutKing wrote on 2021-11-22, 00:30:
That's great news, thanks for the update. […]
Show full quote
trozx wrote on 2021-11-21, 23:19:
Quick update. I just reflowed all IC pins on one of my voodoo2s with success. The card is properly detected and plays games, alb […]
Show full quote

Quick update. I just reflowed all IC pins on one of my voodoo2s with success. The card is properly detected and plays games, albeit the colors are not right. Might have to replace output selector and ramdac, after testing, per this thread:

Successfully fixed a 3DFx Voodoo2 with miscolored output

Going to try reflowing the 2nd one as well.

Mojo.jpg
Tomb1.jpg
Tomb2.jpg

That's great news, thanks for the update.

I might give this a go on my faulty STB Voodoo 2 that is no longer detected.

Care to elaborate on the process you used? Did you add any extra solder?

I just reflowed the existing solder joints with plenty of flux paste; no extra solder.

Those fine pitch parts are temperamental and bridge easily. So, I ended up using a micro chisel type tip with a temp of about 800F. Great results.

I don't want to assume your skill level, but take your time, don't rush it. Make sure you're not tired. Plenty of light and good magnification is a must. At the end, you can use a jewelers loupe or your cell phone zoom to inspect your work - no bridges, shiny fillets.

Here's a picture of the tip I used. About 1.17mm at the end.

Hakko micro chisel.jpg
Filename
Hakko micro chisel.jpg
File size
924.79 KiB
Views
1575 views
File license
Public domain

Reply 57 of 73, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I have begun work again on my bad Diamond Voodoo2 card. Since I started the thread, another Voodoo2 card I had in a system has gone bad. This second card, I am not sure of the brand. I am calling it an Atrend, although I recall being informed it wasn't an Atrend. Nonetheless, that's the name I'm using.

Two_bad_Voodoo2s.JPG
Filename
Two_bad_Voodoo2s.JPG
File size
1014.97 KiB
Views
1239 views
File license
CC-BY-4.0

The output of mojo.exe for a known working Voodoo2 on the same system (AMD K5-PR200):

STB_good.JPG
Filename
STB_good.JPG
File size
284.56 KiB
Views
1239 views
File license
CC-BY-4.0

The output of the nonfunctional Diamond Voodoo2:

Diamond_bad.JPG
Filename
Diamond_bad.JPG
File size
299.14 KiB
Views
1239 views
File license
CC-BY-4.0

The output of the nonfunctional Atrend Voodoo2:

Atrend_bad.JPG
Filename
Atrend_bad.JPG
File size
284.56 KiB
Views
1239 views
File license
CC-BY-4.0

From the mojo.exe results, it looks like the FBI is detected on the Diamond, and only one TMU. However, that TMU shows 0 MB. Perhaps 1 bad TMU and cold joints in the other TMU's RAM? Is TMU 0 the chip near the VGA connector, or is that TMU 1?

The mojo.exe result for the recently bad Atrend looks almost identical to the know good result from the STB, with one exception. STB (good) shows TMU PowerOnSense: 0x8d1, whereas Atrend (bad) shows 0x8d9. Anyone know what this means?

Plan your life wisely, you'll be dead before you know it.

Reply 58 of 73, by feipoa

User metadata
Rank l33t++
Rank
l33t++

The log files for the known good STB and malfunctioning Atrend and Diamond cards are attached here.

Filename
STB-good.txt
File size
1.74 KiB
Downloads
46 downloads
File license
CC-BY-4.0
Filename
Diamond-bad.txt
File size
1.87 KiB
Downloads
46 downloads
File license
CC-BY-4.0
Filename
Atrend-bad.txt
File size
1.74 KiB
Downloads
41 downloads
File license
CC-BY-4.0

Plan your life wisely, you'll be dead before you know it.

Reply 59 of 73, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I just checked all pins on the FBI and TMU's on the Atrend card. They are in contact with the pads and not loose. I then checked all the frame buffer memory to ensure the SOJ's are in contact and they are. I decided to reflow the frame buffer SOJ's just in case, but it didn't help.

When I looked for youtube videos on repairing Voodoo2 cards, every one of them had some visually obvious damage, like cracks on the SOJ joints, FBI non-contacting pins, or bent QFP pins.

On the Atrend, for the most part, if I try to play GLQuake, the screen switches to the 3dfx card, but stays black. Sometimes, if you let it sit there long enough, you may be presented with one of these on the screen:

Atrend_playing_GLQuake_1.JPG
Filename
Atrend_playing_GLQuake_1.JPG
File size
179.24 KiB
Views
1208 views
File license
CC-BY-4.0
Atrend_playing_GLQuake_2.JPG
Filename
Atrend_playing_GLQuake_2.JPG
File size
155.41 KiB
Views
1208 views
File license
CC-BY-4.0

After watching this video, https://www.youtube.com/watch?v=hK80Vq0PYxQ , the guy has made a Glide demo which only uses the FBI as a means to isolate the issue. However, he does not link to his created glide demo.

Plan your life wisely, you'll be dead before you know it.